What does JMP assume when uncontrolled input factors are to be estimated?

[#1]

Hi All,

I've been exploring using the uncontrolled input factors option in JMP. Using a toy example of 5 continuous input factors with 2nd order interactions:

DOE(
	Custom Design,
	{Add Response( None, "Y", ., ., . ), Add Factor( Continuous, -1, 1, "X1", 0 ),
	Add Factor( Continuous, -1, 1, "X2", 0 ),
	Add Factor( Continuous, -1, 1, "X3", 0 ),
	Add Factor( Continuous, -1, 1, "X4", 0 ),
	Add Factor( Continuous, -1, 1, "X5", 0 ), Set Random Seed( 351265972 ),
	Number of Starts( 15868 ), Add Term( {1, 0} ), Add Term( {1, 1} ),
	Add Term( {2, 1} ), Add Term( {3, 1} ), Add Term( {4, 1} ), Add Term( {5, 1} ),
	Add Term( {1, 1}, {2, 1} ), Add Term( {1, 1}, {3, 1} ),
	Add Term( {1, 1}, {4, 1} ), Add Term( {1, 1}, {5, 1} ),
	Add Term( {2, 1}, {3, 1} ), Add Term( {2, 1}, {4, 1} ),
	Add Term( {2, 1}, {5, 1} ), Add Term( {3, 1}, {4, 1} ),
	Add Term( {3, 1}, {5, 1} ), Add Term( {4, 1}, {5, 1} )}
);

20 experiments are recommended as a default. 

When I change one factor to uncontrolled, the recommended experiments remains at 20. So the uncontrolled factor is being treated in exactly the same way as a controllable continuous factor? The JSL reflects this as an uncontrolled factor is shown as just "Add Factor" with no options. 

I understand to correctly alter the number of experiments that the distribution of the uncontrolled factor needs to be known. But is it robust to assume that the number of draws from the uncontrolled factor distribution need to only equal those under controlled factor conditions? 

Kind Regards,

Ben

[#2 · Accepted solution]

The uncontrolled factor type is a continuous factor for which the levels in the experiment are unknown. You will record the levels during each run of the experiment.

The usual Design Evaluation information is not provided because it is computed from the model matrix, which in turn depends on the design.

I am not sure what you mean by, "I understand to correctly alter the number of experiments that the distribution of the uncontrolled factor needs to be known. But is it robust to assume that the number of draws from the uncontrolled factor distribution need to only equal those under controlled factor conditions?" You may select any number of runs at or above the minimum number of runs as always. You just don't have any Design Evaluation reports to help you assess the chosen number of runs.

Note that the default number of runs is based on some simple heuristics. That is all.

[#3]

Thanks for your time on this Mark, the first line answers my question precisely. 

I think potentially there could be a methodology akin to sample size calculations to apply in the case of uncontrolled factors to treat them differently from continuous factors. However, I'm sure this has been looked into already.